The Institute for Security and Technology (IST) recognizes the importance of the Trump Administration’s recently released America’s AI Action Plan, alongside several related Executive Orders. Together, these documents outline the administration’s priorities for domestic and international audiences on key policy issues like accelerating AI adoption, protecting the AI supply chain from adversaries, advancing public-private collaboration, and pursuing AI interpretability and control research. Like many others, IST was pleased to submit comments in response to the Office of Science and Technology Policy’s February request for information in support of the Plan’s development, and are heartened to see addressed many of the key issues that we raised.
While the Plan and accompanying Executive Orders address many critical issues, we note a missed opportunity in adequately addressing the unprecedented future power of AI tools. As emphasized in our February comments, AI represents the foremost national security priority for the United States with the potential emergence of artificial general intelligence (AGI) posing significant challenges. Although debates about AGI’s timeline and likelihood continue, its possible arrival in the near term warrants careful consideration. Overlooking this possibility risks national security both in terms of maintaining strategic advantage and ensuring effective control over such powerful systems. IST encourages the Administration to proactively incorporate this urgent issue into its policymaking efforts.
The first pillar focused on Accelerating AI Innovation sets in motion a range of key activities ranging from enlarging compute access to protecting U.S. AI innovation. IST supports the plan’s emphasis on preventing model theft and adversarial misuse while taking important steps to accelerate AI adoption across the national security enterprise. This includes initiating structured assessments of AI adoption—comparing progress across U.S., allied, and adversarial defense establishments—to ensure a U.S. and allied advantage. This is directly in line with work that IST has done to date as part of our Strategic Balancing Initiative, and elsewhere. IST is pleased to highlight our ongoing activities relevant to this pillar.
- IST’s SL5 Task Force. As we approach AGI and the threshold of automated AI R&D, most frontier AI labs remain unprepared to defend against the world’s most capable nation-state actors who might wish to expropriate or manipulate the technology. We are excited to see calls in the Plan for the U.S. government to work more closely with the private sector to understand security risks and for new technical standards for high-security data centers for the military’s and Intelligence Community’s use. However, private sector frontier AI labs also need such protections. To address this gap—and the parallel risk that the models themselves may pose an insider threat—IST has convened industry stakeholders to enable U.S. AI labs to achieve the necessary Security Level 5 (SL5) posture. This technical working group has developed an initial SL5 standard and is pursuing an industry-focused technical roadmapping to achieve high-security datacenters.
- Integration of AI into U.S. National Security Functions. Since 2017, IST has engaged deeply with the technical community to explore the implications of AI for the battlefield. We maintain ongoing collaboration with leaders across the military, science, national security, and AI industry to identify shared risks and opportunities associated with integrating cutting-edge AI capabilities into key national security dimensions, including Nuclear Command, Control, and Communication (NC3). IST also facilitates strategic foresight efforts by designing and conducting tabletop exercises for stress-testing the strategic, operational, and tactical consequences of AI integration in conflict.
- IST’s AI Risk Reduction Initiative. Reaping the benefits of emerging technologies must be accompanied by efforts to understand and mitigate any resulting risks. As part of a multi-year initiative, IST has been exploring the risks of frontier AI and developing mitigation strategies mapped to the relevant stage of the AI lifecycle. This line of effort began with a study on the risks posed by “open access” foundation models, and was followed by additional phases resulting in recommendations on improving AI interpretability and explainability and strategies for mitigating the malicious use of AI. In our latest ongoing phase of work, IST is researching and convening working group discussions on mitigating AI loss of control.
We note a tension within this pillar’s calls for protecting and advancing U.S. AI innovation, investing in AI control, and prioritizing open-source and open-weight AI models. While the open-source software movement has driven software quality, access, innovation, and cybersecurity, openness in software is not directly comparable to openness in AI model weights—they are fundamentally different. Today’s frontier AI models may not yet require strict security or containment measures, but rapid progress suggests a future where controlling these powerful models involves risks akin to nuclear proliferation, with some sectors like biosecurity already facing such challenges. IST remains committed to fostering innovation while developing prudent risk management strategies. Crucially, there must be an honest assessment of the national security risks posed by openly sharing increasingly powerful AI models. If these tools are freely shared with Chinese labs and military entities, we risk undermining efforts to restrict Chinese access to U.S. AI capabilities. While open source brings many benefits, a serious and coordinated national security discussion on the limits of openness is urgently needed. IST looks forward to engaging in and potentially helping lead this critical dialogue.
The plan’s second pillar on Building American AI Infrastructure highlights the essential ingredients of America’s AI future—specifically, compute capacity and the energy to fuel it. At the same time, this infrastructure must be secure by design, from the start. IST highlights the following ongoing or planned activities relevant to this pillar.
- IST’s Secure-by-Design work. With IST’s emerging focus on driving the Secure by Design movement forward, a key element will be the twin exploration of both how to ensure that the software that powers AI is secure-by-design and how to use this powerful technology to improve software security, such as eliminating entire classes of coding error.
- Energy Security & Resilience. The accompanying Executive Orders take the first steps in advancing the Plan’s energy goals by enhancing permitting processes, incentives, and land access. Nevertheless, affordable power will remain a near- to mid-term constraint. To address this, IST will soon announce a new Energy Security & Resilience Initiative to guide both immediate actions and long-term planning for a stable, cost-effective energy supply needed for AI expansion and the nation’s broader economic success. This initiative will complement regulatory efforts outlined in the Executive Orders with technical guidance on topics such as project siting, adopting grid-enhancing technologies (e.g., dynamic line ratings, flexible AC systems, power flow control), and implementing safeguards to protect new infrastructure—ensuring a resilient energy supply that supports growth without burdening local communities.
- Cybersecurity for Distributed Energy Resources. As AI infrastructure expands, distributed energy resources (DERs) are becoming critical to the U.S. energy supply, enabling more cost-effective and reliable power. As the Plan requires “those gaining access to federal lands to build compute and its accompanying energy infrastructure…[to] meet certain baseline security requirements,” similar measures are needed to address the energy sector’s expanding attack surface nationwide. IST and our partners are convening U.S. industry, researchers, and policymakers to advance practical cybersecurity standards, enhancing the resilience of DER systems and the broader grid.
The plan’s third pillar on Leading in International AI Diplomacy and Security acknowledges that U.S. success in global AI competition depends on strong, consistent international partnerships. While it rightly emphasizes collaborating with allies to uphold American values against authoritarian rivals, some actions risk appearing zero-sum. IST strongly supports tightening export controls—particularly on advanced computing and semiconductors—to limit the PRC’s military modernization and frontier AI capabilities. This strategy depends on international partners, many of whom may face short-term economic costs from restricted sales. We urge the Administration to balance these costs with opportunities for partners to benefit from trade and cooperation with the United States and like-minded allies. IST is pleased to highlight our ongoing activities relevant to this pillar and its implementation.
- IST’s AI Chip Export Control Initiative. While the United States and allied nations have implemented export controls to limit adversaries’ access to powerful microprocessors, persistent compliance gaps and loopholes undermine these efforts, threatening U.S. technological leadership and national security. To address these vulnerabilities, IST launched a comprehensive investigation into the root causes of export control compliance failures, which will lead to recommendations for an enhanced multi-agency AI chip export controls enforcement program to close these critical gaps. The Plan notes these same challenges, and IST looks forward to contributing to the solution set.
- IST’s U.S.-China Deep Tech Investment Research. Recognizing the critical role of venture capital in sustaining U.S. technological leadership, particularly in hardware-intensive and dual-use AI technologies, IST is undertaking a detailed comparative analysis of venture capital strategies in the United States and PRC. This initiative, which builds on IST’s previous work on geopolitical competition with China, assesses investment priorities, incentives, and regulatory barriers, mapping how private sector investment aligns with national security objectives.
IST predicts that this decade will end much differently than it began, in large part driven by AI. Rapid change will test our society, international systems, and perhaps global stability. In order to meet these challenges, the federal government must invest in the people, tools, and partnerships to build a secure, innovative, and strategically-balanced global AI ecosystem. IST stands ready to support the Administration, likeminded nations, and the private sector in pursuing this future.
